ITT: We discuss processors for Steam & whether Westmere 2C/4T should be resurrected?

[cbn]

Its at 32nm for capacity reasons.

Well, if capacity is an issue (even in the future) then I say that is another good reason for 2C/4T Westmere.

See the thing is, It doesn't have to be the ultimate processor. It doesn't need to be the most bleeding edge tech. It just needs to be a great value, the classic superior processor design on an old node that beats inferior processor designs (for desktop) like atom that are on advanced nodes.

 

[cbn]

Hmm..Westmere-EP is Clarkdale?
wut?!
Idk..but where I come from..
This is Westmere-EP:
http://valid.canardpc.com/ev925l
and this is Clarkdale:
http://valid.canardpc.com/2861123

This is the only 2-core Westmere-EP: http://www.cpu-world.com/CPUs/Xeon/Intel-Xeon%20X5698%20-%20AT80614007314AA.html

ex: http://valid.canardpc.com/lmfipi

Oh, I see they also made a dual core based on the Westmere hexcore LGA 1366 die.

 

[witeken]

Well, if capacity is an issue (even in the future) then I say that is another good reason for 2C/4T Westmere.

See the thing is, It doesn't have to be the ultimate processor. It doesn't need to be the most bleeding edge tech. It just needs to be a great value, the classic superior processor design on an old node that beats inferior processor designs (for desktop) like atom that are on advanced nodes.

There is nothing a Westmere CPU can do that a Broadwell CPU at low frequency can't do.

 

[cbn]

There is nothing a Westmere CPU can do that a Broadwell CPU at low frequency can't do.

Of course that is true.

But Broadwell is going to cost a lot more money than 2C/4T Westmere + iGPU or a desktop atom.

I want to know about alternatives to known budget desktop processors (like atom), and I think Westmere will probably fit the bill much better than Broadwell.

 

[witeken]

Broadwell isn't going to cost a meaningful amount more. There will be Broadwell Celerons for 40 bucks. Isn't that good enough?

Intel also had Atom to compete with ARM.

You are searching for problems where there aren't. The challenge isn't how to get ARM out of the desktop, the challenge is how to get into the Android smartphone and tablet.

 

[cbn]

Broadwell isn't going to cost a meaningful amount more. There will be Broadwell Celerons for 40 bucks. Isn't that good enough?

After factoring in the motherboard costs, BGA atom will still be cheaper. I'm guessing the comparison in price is somewhere around $85 for the Celeron vs. $50 for BGA atom if using lowest end motherboards as the starting point.

Intel also had Atom to compete with ARM.

For the most entry level "desktop" (in the form of highly integrated Android set-top boxes, that I mentioned earlier in the thread could be converted to SteamOS for streaming Windows games), I would think Z3735F vs. Rockchip RK3288 (1.8 GHz Cortex A17 quad core) would be a good comparison.

Unfortunately I think even the Rockchip RK3288 is going to be faster if these Geekbench test results of Cortex A12 vs. Z3745 are to be taken seriously:

http://browser.primatelabs.com/geekbench3/compare/734687?baseline=673456

NOTE: ARM claims Cortex A17 is a little over 10 % performance uplift compared to the Cortex A12 used in the benchmark according to this article

NOTE: The benchmarked Bay Trail Z3745 has 17.1 GB/s memory bandwidth compared to 10.6 GB/s for the Z3735F we'll probably commonly find in Bay Trail set-top boxes.

Then, of course, there is Apple at the high end of custom ARM cores. If they begin to pave out an ecosystem for ARM at the higher end performance, we should begin to see a migration of new desktop type ARM processor based apps to Android, etc.

You are searching for problems where there aren't. The challenge isn't how to get ARM out of the desktop, the challenge is how to get into the Android smartphone and tablet

Here is what I wrote in post #43

To put things very bluntly and succinctly, I think Apple may very well catch Intel off guard with respect to desktop performance with ARM.

Take Apple's Cyclone CPU core or one its successors and put it an form factor that is not thermally constrained (eg, Apple TV and make it it more like a Mac Mini), boost clocks and we could have whole wave of desktop like Apps we are not used to seeing in the ARM ecosystem follow soon afterward.

Therefore, I believe Intel needs to approach and think about value desktop in a way they have never thought about before.

Once ARM catches Intel off guard, it will be very difficult for Intel to regain what they lost IMO (maybe a good example of this is the phone SOCs). Instead, I would like to see Intel take some kind of pro-active and aggressive stance now than have to react defensively later on.

P.S. Also at some point, I think Intel needs to plan on offering eMMC 5.0/UFS 2.0 or one its successors as an option for their big core APUs. This way the cheapest class of big core motherboards can have a BOM lowering form of primary storage if necessary

 

[witeken]

Once ARM starts to take some market share, Intel will already have Goldmont & Cannonlake.

 

[cbn]

Not sure what is going to happen with Goldmont or Cannolake, but for some types of full desktop TDP Intel big cores (like 2C/4T Westmere + on package iGPU) I would like to see a pricing structure that encourages BOM lowering components on the motherboard. Some examples would be soldered on RAM and some type of fast type of flash (eMMC 5.0, UFS 2.0, etc.). This, in addition to SATA ports so conventional storage can also be used.

Maybe even use DC-in power (with AC adapter brick) in some cases.

This way the total cost of the budget desktop/PC based gaming console/HTPC can drop in ways beyond simply lowering the price of the processor and PCH.

 

[cbn]

You can also make a dualcore Haswell or Broadwell without IGP, memory controller and so on.

If Haswell was made like the Westmere you link, it would be what, 50mm2? Broadwell, 30mm2?

Revisting this idea of dual core Haswell or Broadwell without memory controller and iGPU (which Intel doesn't make at the moment):

What would be the best way of going about doing this? How do we get Intel to make these chips ShintaiDK is referring to?

Would the best way be to design some type of on package iGPU with memory controller that is reusable across different cpu uarchs? (Remember the original Clarkdale on package graphics used a memory controller from the LGA 775 era of chips. So Intel was able to stretch things across a new cpu uarch even with a memory controller meant for FSB)

Maybe this initial graphics/memory controller on package chip could be designed for Westmere first with the idea of using it later on with the dual core Haswell and Broadwell Chips ShintaiDK mentioned?

 

[TuxDave]

How do we get Intel to make these chips ShintaiDK is referring to?

Easy. Buddy up with an OEM or startup and convince them that this will be a great idea and get them to commit to buying millions of these chips. Then give a call to Intel waving all that $$$. The finance guys will determine it's worth their time to dedicate resources to and you'll have your chips.

 

[cbn]

Deleted.

 

[cbn]

Easy. Buddy up with an OEM or startup and convince them that this will be a great idea and get them to commit to buying millions of these chips. Then give a call to Intel waving all that $$$. The finance guys will determine it's worth their time to dedicate resources to and you'll have your chips.

......So with that idea being so expensive and potentially competing with Intel's current line-up excessively at this time, I would have to imagine Intel would release a dual core Haswell without memory controller and iGPU in 2016 at the soonest. And more likely 2017 and beyond.

Furthermore, such a chip would have to fill a gap Intel is either not willing to fill or can't fill well enough with the current technology of 2016 or 2017+.

With that mentioned, I think a better situation already exists today with 2C/4T Westmere. Sure its 32nm, but it is still very very cheap for Intel to make and it doesn't compete with their existing line-up to any significant degree.

I mostly see Westmere 2C/4T replacing Braswell (for desktop, not laptop) in a few areas outside of niches like fanless desktop. But mostly I see Westmere 2C/4T + iGPU going beyond where Braswell for desktop is capable of going. An area that Intel doesn't have yet.

 

[cbn]

Regarding profit to Intel from Westmere 2C/4T and the 65nm H55/H57 PCH, I don't expect this to be very high. (Fortunately, these older Intel designs have already paid for themselves though so hopefully it is workable to those at Intel responsible for making these decisions)

However, I do think it would help establish Intel in new distribution channels.

For example, If Intel sold Westmere 2C/4T and the 65nm PCH to Rockchip, I would expect to see this big core product (with a new iGPU/memory controller) in the hands of Tronsmart, Rikomagic and a plethora of small chinese companies. These companies I am speaking of normally specialize in small ARM powered set-top boxes (because while SOC is a poor value for non-mobile usage, it is the only affordable form of tech that is available to them).

Instead of being stuck with SOC for these entry level forms of desktops, I would like to see them use something much more powerful and a better value. This so they can get more bang for the buck in their attempts at making entry level PC consoles.



For Rockchip I think this would be a win as well because it would allow them to focus more on putting chips like RK3288 in tablets rather than splitting their efforts between mobile and non-mobile applications. (Maybe HDMI stick being the exception at the extreme low end.)